Research

I have now completed my PhD in the Computer Architecture Group, supervised by Dr. Simon Moore. My dissertation was on Rentian Locality in Chip Multiprocessors. The very high-level summary is that computational complexity isn't what really matters anymore - as communication costs dominate over computational ones. Moreover as we scale closer to the limits of physics, old models and strategies (from the High Performance Computing domain) for dealing with this communication cost no longer apply. This thesis provides new fundamental insights into the complexity of communication in computation, describes how to model and exploit locality, and what it means for future software and computer architectures:

We are entering an exciting new time in computing, where communication costs dominate over the raw computational costs. Old assumptions and models of computational complexity that assume cheap communication need to be seriously revisited in an era when moving a word of data from one part of a chip to another can consume a hundred times the energy of 32-bit arithmetic operations. This thesis argues that it is the communication costs of algorithms, rather than their computation costs, that will dominate future computing concerns. That, as we move to thousands of cores on a chip, the physical spatial locality of computation and data becomes critical to performance and cost. However, there is very little in the way of theory, models, or even characterisation of such locality for Chip-Multiprocessors (CMP). This thesis adapts and extends the existing theory and models of wire locality in VLSI circuits to the physical and temporal locality of software running on CMPs. It aims to provide a new foundation for characterising, modelling, predicting and exploiting the communication properties of software, which as we show, exhibits Rentian fractal scaling. In doing so, it lays a new communication-centric foundation for CMP software and hardware, and provides fundamental insights into their continued technological scaling.

I also examined whether the type of locality seen in computer circuits also exists in biological and other systems - including a collaboration with the Brain Mapping Unit regarding the human brain which attracted some press including here.

Immediately after completing the dissertation I founded a high-tech startup (Fonleap), and thus have not had any time to publish some of the most far-reaching portions of this dissertation as of yet, however the earlier parts can be found in the publication list below:

D. Greenfield, A. Banerjee, J.G. Lee and S.W. Moore, Implications of Rent's Rule for Network-on-Chip Design and its Fault Tolerance, NOCS 2007
S.W. Moore, D. Greenfield, The Next Resource War: Computation vs. Communication, SLIP 2008
D. Greenfield, S.W. Moore, Fractal Communication in software data dependency graphs, SPAA 2008
D. Greenfield, S.W. Moore, Implications of Electronics Technology Trends to Algorithm Design, BCS Computer Journal, 2009 (based on publication at BCS Visions of Computer Science Conference 2008)
D. Bassett, D. Greenfield (joint first author), A Meyer-Lindenberg, D. Weinberger, S.W. Moore, E Bullmore, Efficient Physical Embedding of Topologically Complex Information Processing Networks in Brains and Computer Circuits, PLoS Computational Biology 2010

Background

(Apologies in advance if this sounds a bit like a CV)

I have been working and consulting for high-tech startups since entering high-school. I have worked for various Silicon Valley companies in architecture, design and team-lead roles. This includes designing next-generation graphics chips at SGI, architecting and designing SIMD media processors at a startup, designing chips for Gigabit IPSec and TCP/IP stacks at another startup (acquired by nVidia). I have also designed cutting-edge software as a contractor for companies in Australia.

From the University of New South Wales, I have a Bachelor of Computer Engineering degree as well as a Masters by Research degree in Systems Biology on New and Hybrid Methods for Simulating Biochemical Systems. As an undergraduate, I was placed into accelerated programs, taking additional graduate level subjects in Mathematics, Physics and Computer Science (including on exchange at the University of Illinois, Urbana-Champaign), whilst exempted from some first year subjects. As a result, combined with my work experience, I draw from a very wide field of expert knowledge, and enjoy cross-pollinating ideas across multiple disciplines. Additionally, I have given class lectures and invited talks, including presenting my research in industry, such as at ARM and Samsung HQ.

While my current focus is on future Computer Architectures, I have an insatiable apetite for mastering new things, and I maintain an interest in many other areas including Bioinformatics (mainly Systems Biology), Statistical Machine Learning, Sustainable Energy and Development.

I have won various awards and scholarships including the Gates Cambridge Scholarship, lead teams to become state champion of ACS and ACM programming competitions, and represented Australia at an international level. I have been active in many clubs, societies and humanitarian organisations typically in a leadership or even founding role.

Contact

Daniel Greenfield
University of Cambridge
15 JJ Thomson Avenue
Cambridge CB3 0FD
United Kingdom

Email: Daniel.Greenfield (at) cl.cam.ac.uk

Personal

Some photos can be found here